Lineage Switch As a Relapse Mechanism of Pre-B Acute Lymphoblastic Leukemia Following CD19 CAR

Resistance or relapse following adoptive immunotherapy can occur as a result of antigen loss. T cells armed with chimeric antigen receptors (CARs) targeting CD19 have shown remission rates of over 70% in early clinical trials, although reports of relapse in the presence of persistent CAR T cells are emerging. To further investigate potential resistance mechanisms to CAR T cell therapy, we used 2 syngeneic murine pre-B acute lymphoblastic leukemia (ALL) cell models driven by E2a:PBX1 and Eu-RET. Transplantable cell lines generated from both models have high expression of CD19, CD22, B220, CD43 and CD127, with absence of surface immunoglobulins consistent with a pre-B cell phenotype and result in lethality within 21 days.  Treatment with CD19 CAR T cells induced long-term remission with persistence of CAR T cells beyond 100 days. However, mice bearing E2a:PBX1-driven leukemia experienced relapses occurring in the majority of CAR treated animals within 1 year, all with loss of CD19 expression.  Relatively early post-CAR relapses retained a pre-B phenotype with isolated loss of CD19 extracellular expression by flow cytometry, loss in CD19 exon 1 and 2 mRNA but intact mRNA for all other CD19 exons. In contrast, late post-CAR relapses demonstrated complete loss of CD19 protein and mRNA expression with concomitant loss of the major B cell transcription factors PAX5 and EBF1 suggesting loss of the B-cell developmental program. Furthermore, late post CD19 CAR relapses typically gained myeloid, stem cell or T cell phenotypic markers, consistent with a lineage switch, which was confirmed at the genomic level by RNA-seq of multiple late relapse samples.  We could not identify cells in leukemia culture lacking pre-B phenotype either by single cell cloning following depletion of CD19+ cells indicating that cells with the late relapse phenotype do not exist as a rare population. Finally, we could also demonstrate intermediate phenotypes of post-CD19 CAR relapse in vivo with co-expression of both myeloid (Gr1, CD11b) and B cell markers (B220, CD22) on the same cells, suggesting a differentiation rather than a selection process. We next undertook serial in vivo passaging experiments, and confirmed leukemic-initiating potential of all CD19-loss relapses irrespective of relapse phenotype. Interestingly, leukemic relapse with CD19 expression loss that retain a B-cell program rapidly regain CD19 upon in vivo passage in the absence of CD19 CAR pressure. However, relapses due to lineage reprogramming retained a stable myeloid phenotype upon serial passage without regain of CD19 or other B cell markers. Knock-out of CD19 from leukemic cells using the CRISPR/CAS system did not alter proliferative capability or phenotype during prolonged (30 day) culture and did not diminish engraftment capacity and lethality in vivo, confirming that loss of CD19 alone is insufficient to drive lineage switch. Overall, we demonstrate a novel mechanism of resistance to immunotherapy via lineage switch and demonstrate reprograming potential of ALL under lineage-selective pressure. This observation was repeatedly seen in one murine model (E2a:PBX1) but not in another (Eu-RET), suggesting that the potential for this to occur may depend on the genetic subtype and at what stage the initiating mutation occurred. These results have important clinical implications for the development of antigen-targeting therapies for ALL and, potentially, other cancers.